Four hours of silence
A company that looks after the IT of 14 client firms (an MSP, a managed service provider) runs a night shift of one person. On a Friday night the program that collected status checks lost its network connection at 01:10. Nothing crashed. The dashboard simply stopped receiving news and, with no new red to show, kept showing the last green it knew.
At 05:30 one client’s online payment system had been down for three hours. The dashboard still said green. The night engineer had looked at it twice and seen no reason to look closer. The outage was real, the monitoring was blind, and the screen was lying by leaving things out.
Nobody did anything wrong that night. The tool did exactly what it was built to do: show the last status it had. The problem is that “the last status I had” and “the status right now” look identical on a green tile.
Silence should look like silence
AverDesk’s availability view has two parts: the current status of everything being watched, and a bar of 90 small cells, one per day. The rule for that bar is simple and slightly stubborn. A day with no recorded checks is a gap. It is drawn as a gap. It never turns green just because nothing bad was reported, because nothing at all was reported.
This sounds like a small design choice. In practice it changes how the night shift reads the screen. A row of green with a grey hole in it is a question. Why is there a hole? Did the helper program on the server stop? Did someone switch the check off? Did the server move? Any of those is worth five minutes, and five minutes at 01:30 is much cheaper than three hours at 05:30.
The same honesty applies when you group things. A group takes the worst status of its members. If one server in a cluster is red, the cluster is red. If one is unknown, the group is not green either. You cannot average your way to a calm dashboard.
Planned versus what actually happened
Maintenance is the other place where status screens cheat. Work is announced for Saturday from 02:00 to 04:00, the engineers actually finish at 05:15, and Monday’s report shows a tidy two-hour planned outage.
AverDesk keeps the announced window and the actual window as two separate records. You plan one, you log the other, and the difference is visible. On the estate map planned work shows as an orange ring and an outage as a red one, so a glance tells you whether what you are looking at was expected.
Because planned work is attached to a specific thing on the map, its effect travels along the links. Announce maintenance on a storage system, and every service that runs on servers attached to that storage shows the window too. The maintenance view lets you see that list before you send the announcement. That is a good moment to discover that the client’s payment system is on it.
The right signal to the right person
Showing the truth is only half the job. Someone has to hear about it, and not everybody should hear about everything. AverDesk sends notifications inside the app, by email or through Telegram. You choose who gets what at five levels: the whole platform, your subscription, one organization, one team or one person. The night engineer can follow one client’s critical services on Telegram and leave the rest to email, while the client’s own team sees only their own organization.
Repeated identical alerts are muted. If a server flips between red and green for an hour, you are told once that it is flapping, not sixty times. People who are woken sixty times stop reading messages, and then the sixty-first one is the one that mattered.
What the night shift does differently now
After that Friday the company changed one habit. The first thing the night engineer does at the start of a shift is scan the 90-day bars for today’s cell. Not for red. For grey. A gap at 22:00 means a collector or a server helper needs attention before anything else, because for the rest of the night that gap is where the surprises will hide.
It is a small routine, and it depends entirely on the tool refusing to guess. If your current dashboard has never shown you a gap, it is worth asking whether that is because there has never been one.
You can see the availability view on the status page, or start with the Free plan and point one helper program at one server. The first gap you see will probably teach you something.
Read next
Start with one service
The fastest way to fail at an inventory is to import everything on day one. Here is how to start with a single service and grow from something that already works.
· 5 min read
What AverDesk is and who it is for
A plain description of the product: a list of everything your company runs, a map of what depends on what, and a help desk, all in one place.
· 6 min read